How to choose a UX agency
Start by writing down the question you need answered, then pick the type of firm whose main method answers it. After that, judge candidates on the evidence in their case studies and on who will act on the findings. The checks below are in the order we would run them.
Define the question before the shortlist
UX agencies differ more by method than by quality. A research consultancy and a service design firm will both do good work on the wrong brief.
| Your situation | Type of firm | Firms in this guide |
|---|---|---|
| We do not know why users drop out or complain | Research-led consultancy | Nielsen Norman Group, AnswerLab, Valsplat |
| Our online shop converts below expectations | Research-led, e-commerce benchmark | Baymard Institute |
| We need to decide what the product should become | Product strategy and UX studio | Clearleft, Blink |
| We know the problem and need it designed and built | UX-plus-engineering studio | BlackCard, Ergosign |
| The issue spans channels, staff and back-office processes | Service design firm | Livework, frog |
Evidence that an agency does real research
- Case studies describe the method: who was interviewed or tested, how many people, and what was asked
- Findings are shown, including ones that contradicted the client's assumptions
- The story links a finding to a design decision and then to a measured result
- Named researchers exist on the team page, separate from designers
- The firm publishes articles, reports or talks about its methods
- It can explain how it recruits participants and screens them
Nine questions for the first conversation
- Which method would you use for our question, and why that one?
- Who recruits participants, and how do you make sure they match our users?
- Who will run the sessions, and can we observe them live?
- What will the deliverable look like? Can we see an anonymised example?
- How do you prioritise findings, and do you estimate effort to fix them?
- How will you use our existing analytics and support data?
- What happens after the report: do you design, build, or hand over?
- Who owns the recordings and raw data, and how is personal data handled under GDPR?
- What would make you advise us not to run this project?
Warning signs
- Personas and journey maps offered without any plan to talk to users
- A proposal that names the solution before the research has started
- Reports measured in page count instead of decisions
- No access to session recordings or raw notes
- A team that changes completely between the pitch and the kick-off
Decide who will act on the findings
Research has value only when something changes afterwards. Before you sign, name the team that will turn recommendations into design and code, and involve it from the first session.
There are three workable setups. A research firm reports to your in-house product team. A strategy studio hands tested designs to your developers. Or a UX-plus-engineering studio does the whole chain, which suits companies without a product team of their own.
Look at structure and ownership
Ownership shapes how a firm sells and staffs work. Blink has belonged to Mphasis since 2021, and frog is part of Capgemini Invent. Clearleft is employee-owned, and Livework describes itself as independent. AnswerLab widened its positioning from research to experience strategy in February 2026.
None of these facts is good or bad on its own. A parent company can bring reach and delivery capacity. An independent firm may have fewer reasons to recommend a large follow-on programme.
Begin with a contained first project
A packaged review or one test round is a low-risk trial. You see how the firm plans, how it treats participants and how clearly it reports.
Judge the result by one test: did the team learn something it did not know, and is it obvious what to do next?
Common questions
Does a UX agency need to know my industry?
Domain knowledge shortens the start, especially in regulated or technical fields. Method matters more. A firm that recruits the right participants and asks good questions will learn your domain quickly, while industry experience cannot rescue weak research.
Should the same agency do research and design?
It is efficient, because knowledge stays in one team. The risk is that a firm grades its own work. If you combine them, ask for at least one test round run by someone who did not design the solution.